[net-next-2.6.git] / net / dccp / ackvec.c

/*
 *  net/dccp/ackvec.c
 *
 *  An implementation of the DCCP protocol
 *  Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
 *
 *      This program is free software; you can redistribute it and/or modify it
 *      under the terms of the GNU General Public License as published by the
 *      Free Software Foundation; version 2 of the License;
 */

#include "ackvec.h"
#include "dccp.h"

#include <linux/dccp.h>
#include <linux/init.h>
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/slab.h>

#include <net/sock.h>

static kmem_cache_t *dccp_ackvec_slab;
static kmem_cache_t *dccp_ackvec_record_slab;

static struct dccp_ackvec_record *dccp_ackvec_record_new(void)
{
	struct dccp_ackvec_record *avr =
			kmem_cache_alloc(dccp_ackvec_record_slab, GFP_ATOMIC);

	if (avr != NULL)
		INIT_LIST_HEAD(&avr->dccpavr_node);

	return avr;
}

static void dccp_ackvec_record_delete(struct dccp_ackvec_record *avr)
{
	if (unlikely(avr == NULL))
		return;
	/* Check if deleting a linked record */
	WARN_ON(!list_empty(&avr->dccpavr_node));
	kmem_cache_free(dccp_ackvec_record_slab, avr);
}

static void dccp_ackvec_insert_avr(struct dccp_ackvec *av,
				   struct dccp_ackvec_record *avr)
{
	/*
	 * AVRs are sorted by seqno. Since we are sending them in order, we
	 * just add the AVR at the head of the list.
	 * -sorbo.
	 */
	if (!list_empty(&av->dccpav_records)) {
		const struct dccp_ackvec_record *head =
					list_entry(av->dccpav_records.next,
						   struct dccp_ackvec_record,
						   dccpavr_node);
		BUG_ON(before48(avr->dccpavr_ack_seqno,
				head->dccpavr_ack_seqno));
	}

	list_add(&avr->dccpavr_node, &av->dccpav_records);
}

int dccp_insert_option_ackvec(struct sock *sk, struct sk_buff *skb)
{
	struct dccp_sock *dp = dccp_sk(sk);
	struct dccp_ackvec *av = dp->dccps_hc_rx_ackvec;
	int len = av->dccpav_vec_len + 2;
	struct timeval now;
	u32 elapsed_time;
	unsigned char *to, *from;
	struct dccp_ackvec_record *avr;

	if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
		return -1;

	dccp_timestamp(sk, &now);
	elapsed_time = timeval_delta(&now, &av->dccpav_time) / 10;

	if (elapsed_time != 0 &&
	    dccp_insert_option_elapsed_time(sk, skb, elapsed_time))
		return -1;

	avr = dccp_ackvec_record_new();
	if (avr == NULL)
		return -1;

	DCCP_SKB_CB(skb)->dccpd_opt_len += len;

	to    = skb_push(skb, len);
	*to++ = DCCPO_ACK_VECTOR_0;
	*to++ = len;

	len  = av->dccpav_vec_len;
	from = av->dccpav_buf + av->dccpav_buf_head;

	/* Check if buf_head wraps */
	if ((int)av->dccpav_buf_head + len > DCCP_MAX_ACKVEC_LEN) {
		const u32 tailsize = DCCP_MAX_ACKVEC_LEN - av->dccpav_buf_head;

		memcpy(to, from, tailsize);
		to   += tailsize;
		len  -= tailsize;
		from = av->dccpav_buf;
	}

	memcpy(to, from, len);
	/*
	 *	From RFC 4340, A.2:
	 *
	 *	For each acknowledgement it sends, the HC-Receiver will add an
	 *	acknowledgement record.  ack_seqno will equal the HC-Receiver
	 *	sequence number it used for the ack packet; ack_ptr will equal
	 *	buf_head; ack_ackno will equal buf_ackno; and ack_nonce will
	 *	equal buf_nonce.
	 */
	avr->dccpavr_ack_seqno = DCCP_SKB_CB(skb)->dccpd_seq;
	avr->dccpavr_ack_ptr   = av->dccpav_buf_head;
	avr->dccpavr_ack_ackno = av->dccpav_buf_ackno;
	avr->dccpavr_ack_nonce = av->dccpav_buf_nonce;
	avr->dccpavr_sent_len  = av->dccpav_vec_len;

	dccp_ackvec_insert_avr(av, avr);

	dccp_pr_debug("%s ACK Vector 0, len=%d, ack_seqno=%llu, "
		      "ack_ackno=%llu\n",
		      dccp_role(sk), avr->dccpavr_sent_len,
		      (unsigned long long)avr->dccpavr_ack_seqno,
		      (unsigned long long)avr->dccpavr_ack_ackno);
	return 0;
}

struct dccp_ackvec *dccp_ackvec_alloc(const gfp_t priority)
{
	struct dccp_ackvec *av = kmem_cache_alloc(dccp_ackvec_slab, priority);

	if (av != NULL) {
		av->dccpav_buf_head	= DCCP_MAX_ACKVEC_LEN - 1;
		av->dccpav_buf_ackno	= DCCP_MAX_SEQNO + 1;
		av->dccpav_buf_nonce = av->dccpav_buf_nonce = 0;
		av->dccpav_time.tv_sec	= 0;
		av->dccpav_time.tv_usec	= 0;
		av->dccpav_vec_len	= 0;
		INIT_LIST_HEAD(&av->dccpav_records);
	}

	return av;
}

void dccp_ackvec_free(struct dccp_ackvec *av)
{
	if (unlikely(av == NULL))
		return;

	if (!list_empty(&av->dccpav_records)) {
		struct dccp_ackvec_record *avr, *next;

		list_for_each_entry_safe(avr, next, &av->dccpav_records,
					 dccpavr_node) {
			list_del_init(&avr->dccpavr_node);
			dccp_ackvec_record_delete(avr);
		}
	}

	kmem_cache_free(dccp_ackvec_slab, av);
}

static inline u8 dccp_ackvec_state(const struct dccp_ackvec *av,
				   const u8 index)
{
	return av->dccpav_buf[index] & DCCP_ACKVEC_STATE_MASK;
}

static inline u8 dccp_ackvec_len(const struct dccp_ackvec *av,
				 const u8 index)
{
	return av->dccpav_buf[index] & DCCP_ACKVEC_LEN_MASK;
}

/*
 * If several packets are missing, the HC-Receiver may prefer to enter multiple
 * bytes with run length 0, rather than a single byte with a larger run length;
 * this simplifies table updates if one of the missing packets arrives.
 */
static inline int dccp_ackvec_set_buf_head_state(struct dccp_ackvec *av,
						 const unsigned int packets,
						 const unsigned char state)
{
	unsigned int gap;
	long new_head;

	if (av->dccpav_vec_len + packets > DCCP_MAX_ACKVEC_LEN)
		return -ENOBUFS;

	gap	 = packets - 1;
	new_head = av->dccpav_buf_head - packets;

	if (new_head < 0) {
		if (gap > 0) {
			memset(av->dccpav_buf, DCCP_ACKVEC_STATE_NOT_RECEIVED,
			       gap + new_head + 1);
			gap = -new_head;
		}
		new_head += DCCP_MAX_ACKVEC_LEN;
	} 

	av->dccpav_buf_head = new_head;

	if (gap > 0)
		memset(av->dccpav_buf + av->dccpav_buf_head + 1,
		       DCCP_ACKVEC_STATE_NOT_RECEIVED, gap);

	av->dccpav_buf[av->dccpav_buf_head] = state;
	av->dccpav_vec_len += packets;
	return 0;
}

/*
 * Implements the RFC 4340, Appendix A
 */
int dccp_ackvec_add(struct dccp_ackvec *av, const struct sock *sk,
		    const u64 ackno, const u8 state)
{
	/*
	 * Check at the right places if the buffer is full, if it is, tell the
	 * caller to start dropping packets till the HC-Sender acks our ACK
	 * vectors, when we will free up space in dccpav_buf.
	 *
	 * We may well decide to do buffer compression, etc, but for now lets
	 * just drop.
	 *
	 * From Appendix A.1.1 (`New Packets'):
	 *
	 *	Of course, the circular buffer may overflow, either when the
	 *	HC-Sender is sending data at a very high rate, when the
	 *	HC-Receiver's acknowledgements are not reaching the HC-Sender,
	 *	or when the HC-Sender is forgetting to acknowledge those acks
	 *	(so the HC-Receiver is unable to clean up old state). In this
	 *	case, the HC-Receiver should either compress the buffer (by
	 *	increasing run lengths when possible), transfer its state to
	 *	a larger buffer, or, as a last resort, drop all received
	 *	packets, without processing them whatsoever, until its buffer
	 *	shrinks again.
	 */

	/* See if this is the first ackno being inserted */
	if (av->dccpav_vec_len == 0) {
		av->dccpav_buf[av->dccpav_buf_head] = state;
		av->dccpav_vec_len = 1;
	} else if (after48(ackno, av->dccpav_buf_ackno)) {
		const u64 delta = dccp_delta_seqno(av->dccpav_buf_ackno,
						   ackno);

		/*
		 * Look if the state of this packet is the same as the
		 * previous ackno and if so if we can bump the head len.
		 */
		if (delta == 1 &&
		    dccp_ackvec_state(av, av->dccpav_buf_head) == state &&
		    (dccp_ackvec_len(av, av->dccpav_buf_head) <
		     DCCP_ACKVEC_LEN_MASK))
			av->dccpav_buf[av->dccpav_buf_head]++;
		else if (dccp_ackvec_set_buf_head_state(av, delta, state))
			return -ENOBUFS;
	} else {
		/*
		 * A.1.2.  Old Packets
		 *
		 *	When a packet with Sequence Number S <= buf_ackno
		 *	arrives, the HC-Receiver will scan the table for
		 *	the byte corresponding to S. (Indexing structures
		 *	could reduce the complexity of this scan.)
		 */
		u64 delta = dccp_delta_seqno(ackno, av->dccpav_buf_ackno);
		u8 index = av->dccpav_buf_head;

		while (1) {
			const u8 len = dccp_ackvec_len(av, index);
			const u8 state = dccp_ackvec_state(av, index);
			/*
			 * valid packets not yet in dccpav_buf have a reserved
			 * entry, with a len equal to 0.
			 */
			if (state == DCCP_ACKVEC_STATE_NOT_RECEIVED &&
			    len == 0 && delta == 0) { /* Found our
							 reserved seat! */
				dccp_pr_debug("Found %llu reserved seat!\n",
					      (unsigned long long)ackno);
				av->dccpav_buf[index] = state;
				goto out;
			}
			/* len == 0 means one packet */
			if (delta < len + 1)
				goto out_duplicate;

			delta -= len + 1;
			if (++index == DCCP_MAX_ACKVEC_LEN)
				index = 0;
		}
	}

	av->dccpav_buf_ackno = ackno;
	dccp_timestamp(sk, &av->dccpav_time);
out:
	return 0;

out_duplicate:
	/* Duplicate packet */
	dccp_pr_debug("Received a dup or already considered lost "
		      "packet: %llu\n", (unsigned long long)ackno);
	return -EILSEQ;
}

#ifdef CONFIG_IP_DCCP_DEBUG
void dccp_ackvector_print(const u64 ackno, const unsigned char *vector, int len)
{
	dccp_pr_debug_cat("ACK vector len=%d, ackno=%llu |", len,
			 		(unsigned long long)ackno);

	while (len--) {
		const u8 state = (*vector & DCCP_ACKVEC_STATE_MASK) >> 6;
		const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;

		dccp_pr_debug_cat("%d,%d|", state, rl);
		++vector;
	}

	dccp_pr_debug_cat("\n");
}

void dccp_ackvec_print(const struct dccp_ackvec *av)
{
	dccp_ackvector_print(av->dccpav_buf_ackno,
			     av->dccpav_buf + av->dccpav_buf_head,
			     av->dccpav_vec_len);
}
#endif

static void dccp_ackvec_throw_record(struct dccp_ackvec *av,
				     struct dccp_ackvec_record *avr)
{
	struct dccp_ackvec_record *next;

	/* sort out vector length */
	if (av->dccpav_buf_head <= avr->dccpavr_ack_ptr)
		av->dccpav_vec_len = avr->dccpavr_ack_ptr - av->dccpav_buf_head;
	else
		av->dccpav_vec_len = DCCP_MAX_ACKVEC_LEN - 1
				     - av->dccpav_buf_head
				     + avr->dccpavr_ack_ptr;

	/* free records */
	list_for_each_entry_safe_from(avr, next, &av->dccpav_records,
				      dccpavr_node) {
		list_del_init(&avr->dccpavr_node);
		dccp_ackvec_record_delete(avr);
	}
}

void dccp_ackvec_check_rcv_ackno(struct dccp_ackvec *av, struct sock *sk,
				 const u64 ackno)
{
	struct dccp_ackvec_record *avr;

	/*
	 * If we traverse backwards, it should be faster when we have large
	 * windows. We will be receiving ACKs for stuff we sent a while back
	 * -sorbo.
	 */
	list_for_each_entry_reverse(avr, &av->dccpav_records, dccpavr_node) {
		if (ackno == avr->dccpavr_ack_seqno) {
			dccp_pr_debug("%s ACK packet 0, len=%d, ack_seqno=%llu, "
				      "ack_ackno=%llu, ACKED!\n",
				      dccp_role(sk), 1,
				      (unsigned long long)avr->dccpavr_ack_seqno,
				      (unsigned long long)avr->dccpavr_ack_ackno);
			dccp_ackvec_throw_record(av, avr);
			break;
		} else if (avr->dccpavr_ack_seqno > ackno)
			break; /* old news */
	}
}

static void dccp_ackvec_check_rcv_ackvector(struct dccp_ackvec *av,
					    struct sock *sk, u64 ackno,
					    const unsigned char len,
					    const unsigned char *vector)
{
	unsigned char i;
	struct dccp_ackvec_record *avr;

	/* Check if we actually sent an ACK vector */
	if (list_empty(&av->dccpav_records))
		return;

	i = len;
	/*
	 * XXX
	 * I think it might be more efficient to work backwards. See comment on
	 * rcv_ackno. -sorbo.
	 */
	avr = list_entry(av->dccpav_records.next, struct dccp_ackvec_record,
			 dccpavr_node);
	while (i--) {
		const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
		u64 ackno_end_rl;

		dccp_set_seqno(&ackno_end_rl, ackno - rl);

		/*
		 * If our AVR sequence number is greater than the ack, go
		 * forward in the AVR list until it is not so.
		 */
		list_for_each_entry_from(avr, &av->dccpav_records,
					 dccpavr_node) {
			if (!after48(avr->dccpavr_ack_seqno, ackno))
				goto found;
		}
		/* End of the dccpav_records list, not found, exit */
		break;
found:
		if (between48(avr->dccpavr_ack_seqno, ackno_end_rl, ackno)) {
			const u8 state = *vector & DCCP_ACKVEC_STATE_MASK;
			if (state != DCCP_ACKVEC_STATE_NOT_RECEIVED) {
				dccp_pr_debug("%s ACK vector 0, len=%d, "
					      "ack_seqno=%llu, ack_ackno=%llu, "
					      "ACKED!\n",
					      dccp_role(sk), len,
					      (unsigned long long)
					      avr->dccpavr_ack_seqno,
					      (unsigned long long)
					      avr->dccpavr_ack_ackno);
				dccp_ackvec_throw_record(av, avr);
				break;
			}
			/*
			 * If it wasn't received, continue scanning... we might
			 * find another one.
			 */
		}

		dccp_set_seqno(&ackno, ackno_end_rl - 1);
		++vector;
	}
}

int dccp_ackvec_parse(struct sock *sk, const struct sk_buff *skb,
		      const u8 opt, const u8 *value, const u8 len)
{
	if (len > DCCP_MAX_ACKVEC_LEN)
		return -1;

	/* dccp_ackvector_print(DCCP_SKB_CB(skb)->dccpd_ack_seq, value, len); */
	dccp_ackvec_check_rcv_ackvector(dccp_sk(sk)->dccps_hc_rx_ackvec, sk,
					DCCP_SKB_CB(skb)->dccpd_ack_seq,
				        len, value);
	return 0;
}

int __init dccp_ackvec_init(void)
{
	dccp_ackvec_slab = kmem_cache_create("dccp_ackvec",
					     sizeof(struct dccp_ackvec), 0,
					     SLAB_HWCACHE_ALIGN, NULL, NULL);
	if (dccp_ackvec_slab == NULL)
		goto out_err;

	dccp_ackvec_record_slab =
			kmem_cache_create("dccp_ackvec_record",
					  sizeof(struct dccp_ackvec_record),
					  0, SLAB_HWCACHE_ALIGN, NULL, NULL);
	if (dccp_ackvec_record_slab == NULL)
		goto out_destroy_slab;

	return 0;

out_destroy_slab:
	kmem_cache_destroy(dccp_ackvec_slab);
	dccp_ackvec_slab = NULL;
out_err:
	DCCP_CRIT("Unable to create Ack Vector slab cache");
	return -ENOBUFS;
}

void dccp_ackvec_exit(void)
{
	if (dccp_ackvec_slab != NULL) {
		kmem_cache_destroy(dccp_ackvec_slab);
		dccp_ackvec_slab = NULL;
	}
	if (dccp_ackvec_record_slab != NULL) {
		kmem_cache_destroy(dccp_ackvec_record_slab);
		dccp_ackvec_record_slab = NULL;
	}
}
Commit	Line	Data
ae31c339 ACM	1	/*
	2	* net/dccp/ackvec.c
	3	*
	4	* An implementation of the DCCP protocol
	5	* Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
	6	*
	7	* This program is free software; you can redistribute it and/or modify it
	8	* under the terms of the GNU General Public License as published by the
	9	* Free Software Foundation; version 2 of the License;
	10	*/
	11
	12	#include "ackvec.h"
	13	#include "dccp.h"
	14
	15	#include <linux/dccp.h>
9b07ef5d ACM	16	#include <linux/init.h>
	17	#include <linux/errno.h>
	18	#include <linux/kernel.h>
ae31c339	19	#include <linux/skbuff.h>
9b07ef5d	20	#include <linux/slab.h>
ae31c339 ACM	21
	22	#include <net/sock.h>
	23
9b07ef5d	24	static kmem_cache_t *dccp_ackvec_slab;
02bcf28c AB	25	static kmem_cache_t *dccp_ackvec_record_slab;
	26
	27	static struct dccp_ackvec_record *dccp_ackvec_record_new(void)
	28	{
	29	struct dccp_ackvec_record *avr =
	30	kmem_cache_alloc(dccp_ackvec_record_slab, GFP_ATOMIC);
	31
	32	if (avr != NULL)
	33	INIT_LIST_HEAD(&avr->dccpavr_node);
	34
	35	return avr;
	36	}
	37
	38	static void dccp_ackvec_record_delete(struct dccp_ackvec_record *avr)
	39	{
	40	if (unlikely(avr == NULL))
	41	return;
	42	/* Check if deleting a linked record */
	43	WARN_ON(!list_empty(&avr->dccpavr_node));
	44	kmem_cache_free(dccp_ackvec_record_slab, avr);
	45	}
	46
	47	static void dccp_ackvec_insert_avr(struct dccp_ackvec *av,
	48	struct dccp_ackvec_record *avr)
	49	{
	50	/*
	51	* AVRs are sorted by seqno. Since we are sending them in order, we
	52	* just add the AVR at the head of the list.
	53	* -sorbo.
	54	*/
	55	if (!list_empty(&av->dccpav_records)) {
	56	const struct dccp_ackvec_record *head =
	57	list_entry(av->dccpav_records.next,
	58	struct dccp_ackvec_record,
	59	dccpavr_node);
	60	BUG_ON(before48(avr->dccpavr_ack_seqno,
	61	head->dccpavr_ack_seqno));
	62	}
	63
	64	list_add(&avr->dccpavr_node, &av->dccpav_records);
	65	}
9b07ef5d	66
ae31c339 ACM	67	int dccp_insert_option_ackvec(struct sock sk, struct sk_buff skb)
	68	{
	69	struct dccp_sock *dp = dccp_sk(sk);
	70	struct dccp_ackvec *av = dp->dccps_hc_rx_ackvec;
	71	int len = av->dccpav_vec_len + 2;
	72	struct timeval now;
	73	u32 elapsed_time;
	74	unsigned char to, from;
02bcf28c AB	75	struct dccp_ackvec_record *avr;
	76
	77	if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
	78	return -1;
	79
ae31c339 ACM	80	dccp_timestamp(sk, &now);
	81	elapsed_time = timeval_delta(&now, &av->dccpav_time) / 10;
	82
2d0817d1 ACM	83	if (elapsed_time != 0 &&
	84	dccp_insert_option_elapsed_time(sk, skb, elapsed_time))
	85	return -1;
	86
	87	avr = dccp_ackvec_record_new();
	88	if (avr == NULL)
	89	return -1;
ae31c339	90
ae31c339 ACM	91	DCCP_SKB_CB(skb)->dccpd_opt_len += len;
	92
	93	to = skb_push(skb, len);
	94	*to++ = DCCPO_ACK_VECTOR_0;
	95	*to++ = len;
	96
	97	len = av->dccpav_vec_len;
	98	from = av->dccpav_buf + av->dccpav_buf_head;
	99
	100	/* Check if buf_head wraps */
02bcf28c AB	101	if ((int)av->dccpav_buf_head + len > DCCP_MAX_ACKVEC_LEN) {
02bcf28c AB	102	const u32 tailsize = DCCP_MAX_ACKVEC_LEN - av->dccpav_buf_head;
ae31c339 ACM	103
	104	memcpy(to, from, tailsize);
	105	to += tailsize;
	106	len -= tailsize;
	107	from = av->dccpav_buf;
	108	}
	109
	110	memcpy(to, from, len);
	111	/*
0e64e94e	112	* From RFC 4340, A.2:
ae31c339 ACM	113	*
	114	* For each acknowledgement it sends, the HC-Receiver will add an
	115	* acknowledgement record. ack_seqno will equal the HC-Receiver
	116	* sequence number it used for the ack packet; ack_ptr will equal
	117	* buf_head; ack_ackno will equal buf_ackno; and ack_nonce will
	118	* equal buf_nonce.
ae31c339	119	*/
02bcf28c AB	120	avr->dccpavr_ack_seqno = DCCP_SKB_CB(skb)->dccpd_seq;
	121	avr->dccpavr_ack_ptr = av->dccpav_buf_head;
	122	avr->dccpavr_ack_ackno = av->dccpav_buf_ackno;
	123	avr->dccpavr_ack_nonce = av->dccpav_buf_nonce;
	124	avr->dccpavr_sent_len = av->dccpav_vec_len;
	125
	126	dccp_ackvec_insert_avr(av, avr);
ae31c339	127
09dbc389	128	dccp_pr_debug("%s ACK Vector 0, len=%d, ack_seqno=%llu, "
ae31c339	129	"ack_ackno=%llu\n",
09dbc389	130	dccp_role(sk), avr->dccpavr_sent_len,
02bcf28c AB	131	(unsigned long long)avr->dccpavr_ack_seqno,
	132	(unsigned long long)avr->dccpavr_ack_ackno);
	133	return 0;
ae31c339 ACM	134	}
ae31c339 ACM	135
7400d781	136	struct dccp_ackvec *dccp_ackvec_alloc(const gfp_t priority)
ae31c339	137	{
9b07ef5d	138	struct dccp_ackvec *av = kmem_cache_alloc(dccp_ackvec_slab, priority);
e4dfd449	139
ae31c339	140	if (av != NULL) {
4a0a50fb	141	av->dccpav_buf_head = DCCP_MAX_ACKVEC_LEN - 1;
02bcf28c	142	av->dccpav_buf_ackno = DCCP_MAX_SEQNO + 1;
ae31c339	143	av->dccpav_buf_nonce = av->dccpav_buf_nonce = 0;
ae31c339 ACM	144	av->dccpav_time.tv_sec = 0;
ae31c339 ACM	145	av->dccpav_time.tv_usec = 0;
4a0a50fb	146	av->dccpav_vec_len = 0;
02bcf28c	147	INIT_LIST_HEAD(&av->dccpav_records);
ae31c339 ACM	148	}
	149
	150	return av;
	151	}
	152
	153	void dccp_ackvec_free(struct dccp_ackvec *av)
	154	{
02bcf28c AB	155	if (unlikely(av == NULL))
02bcf28c AB	156	return;
d5e9b2c7 ACM	157
	158	if (!list_empty(&av->dccpav_records)) {
	159	struct dccp_ackvec_record avr, next;
	160
	161	list_for_each_entry_safe(avr, next, &av->dccpav_records,
	162	dccpavr_node) {
	163	list_del_init(&avr->dccpavr_node);
	164	dccp_ackvec_record_delete(avr);
	165	}
	166	}
	167
9b07ef5d	168	kmem_cache_free(dccp_ackvec_slab, av);
ae31c339 ACM	169	}
	170
	171	static inline u8 dccp_ackvec_state(const struct dccp_ackvec *av,
e4dfd449	172	const u8 index)
ae31c339 ACM	173	{
	174	return av->dccpav_buf[index] & DCCP_ACKVEC_STATE_MASK;
	175	}
	176
	177	static inline u8 dccp_ackvec_len(const struct dccp_ackvec *av,
e4dfd449	178	const u8 index)
ae31c339 ACM	179	{
	180	return av->dccpav_buf[index] & DCCP_ACKVEC_LEN_MASK;
	181	}
	182
	183	/*
	184	* If several packets are missing, the HC-Receiver may prefer to enter multiple
	185	* bytes with run length 0, rather than a single byte with a larger run length;
	186	* this simplifies table updates if one of the missing packets arrives.
	187	*/
	188	static inline int dccp_ackvec_set_buf_head_state(struct dccp_ackvec *av,
	189	const unsigned int packets,
e4dfd449	190	const unsigned char state)
ae31c339 ACM	191	{
ae31c339 ACM	192	unsigned int gap;
a8fc3d8d	193	long new_head;
ae31c339	194
7400d781	195	if (av->dccpav_vec_len + packets > DCCP_MAX_ACKVEC_LEN)
ae31c339 ACM	196	return -ENOBUFS;
	197
	198	gap = packets - 1;
	199	new_head = av->dccpav_buf_head - packets;
	200
	201	if (new_head < 0) {
	202	if (gap > 0) {
	203	memset(av->dccpav_buf, DCCP_ACKVEC_STATE_NOT_RECEIVED,
	204	gap + new_head + 1);
	205	gap = -new_head;
	206	}
7400d781	207	new_head += DCCP_MAX_ACKVEC_LEN;
ae31c339 ACM	208	}
	209
	210	av->dccpav_buf_head = new_head;
	211
	212	if (gap > 0)
	213	memset(av->dccpav_buf + av->dccpav_buf_head + 1,
	214	DCCP_ACKVEC_STATE_NOT_RECEIVED, gap);
	215
	216	av->dccpav_buf[av->dccpav_buf_head] = state;
	217	av->dccpav_vec_len += packets;
	218	return 0;
	219	}
	220
	221	/*
0e64e94e	222	* Implements the RFC 4340, Appendix A
ae31c339 ACM	223	*/
	224	int dccp_ackvec_add(struct dccp_ackvec av, const struct sock sk,
	225	const u64 ackno, const u8 state)
	226	{
	227	/*
	228	* Check at the right places if the buffer is full, if it is, tell the
	229	* caller to start dropping packets till the HC-Sender acks our ACK
	230	* vectors, when we will free up space in dccpav_buf.
	231	*
	232	* We may well decide to do buffer compression, etc, but for now lets
	233	* just drop.
	234	*
0e64e94e	235	* From Appendix A.1.1 (`New Packets'):
ae31c339 ACM	236	*
	237	* Of course, the circular buffer may overflow, either when the
	238	* HC-Sender is sending data at a very high rate, when the
	239	* HC-Receiver's acknowledgements are not reaching the HC-Sender,
	240	* or when the HC-Sender is forgetting to acknowledge those acks
	241	* (so the HC-Receiver is unable to clean up old state). In this
	242	* case, the HC-Receiver should either compress the buffer (by
	243	* increasing run lengths when possible), transfer its state to
	244	* a larger buffer, or, as a last resort, drop all received
	245	* packets, without processing them whatsoever, until its buffer
	246	* shrinks again.
	247	*/
	248
	249	/* See if this is the first ackno being inserted */
	250	if (av->dccpav_vec_len == 0) {
	251	av->dccpav_buf[av->dccpav_buf_head] = state;
	252	av->dccpav_vec_len = 1;
	253	} else if (after48(ackno, av->dccpav_buf_ackno)) {
	254	const u64 delta = dccp_delta_seqno(av->dccpav_buf_ackno,
	255	ackno);
	256
	257	/*
	258	* Look if the state of this packet is the same as the
	259	* previous ackno and if so if we can bump the head len.
	260	*/
	261	if (delta == 1 &&
	262	dccp_ackvec_state(av, av->dccpav_buf_head) == state &&
	263	(dccp_ackvec_len(av, av->dccpav_buf_head) <
	264	DCCP_ACKVEC_LEN_MASK))
	265	av->dccpav_buf[av->dccpav_buf_head]++;
	266	else if (dccp_ackvec_set_buf_head_state(av, delta, state))
	267	return -ENOBUFS;
	268	} else {
	269	/*
	270	* A.1.2. Old Packets
	271	*
0e64e94e GR	272	* When a packet with Sequence Number S <= buf_ackno
	273	* arrives, the HC-Receiver will scan the table for
	274	* the byte corresponding to S. (Indexing structures
ae31c339 ACM	275	* could reduce the complexity of this scan.)
	276	*/
	277	u64 delta = dccp_delta_seqno(ackno, av->dccpav_buf_ackno);
e4dfd449	278	u8 index = av->dccpav_buf_head;
ae31c339 ACM	279
	280	while (1) {
	281	const u8 len = dccp_ackvec_len(av, index);
	282	const u8 state = dccp_ackvec_state(av, index);
	283	/*
	284	* valid packets not yet in dccpav_buf have a reserved
	285	* entry, with a len equal to 0.
	286	*/
	287	if (state == DCCP_ACKVEC_STATE_NOT_RECEIVED &&
	288	len == 0 && delta == 0) { /* Found our
	289	reserved seat! */
	290	dccp_pr_debug("Found %llu reserved seat!\n",
	291	(unsigned long long)ackno);
	292	av->dccpav_buf[index] = state;
	293	goto out;
	294	}
	295	/* len == 0 means one packet */
	296	if (delta < len + 1)
	297	goto out_duplicate;
	298
	299	delta -= len + 1;
7400d781	300	if (++index == DCCP_MAX_ACKVEC_LEN)
ae31c339 ACM	301	index = 0;
	302	}
	303	}
	304
	305	av->dccpav_buf_ackno = ackno;
	306	dccp_timestamp(sk, &av->dccpav_time);
	307	out:
ae31c339 ACM	308	return 0;
	309
	310	out_duplicate:
	311	/* Duplicate packet */
	312	dccp_pr_debug("Received a dup or already considered lost "
	313	"packet: %llu\n", (unsigned long long)ackno);
	314	return -EILSEQ;
	315	}
	316
	317	#ifdef CONFIG_IP_DCCP_DEBUG
	318	void dccp_ackvector_print(const u64 ackno, const unsigned char *vector, int len)
	319	{
84116716 GR	320	dccp_pr_debug_cat("ACK vector len=%d, ackno=%llu \|", len,
84116716 GR	321	(unsigned long long)ackno);
ae31c339 ACM	322
	323	while (len--) {
	324	const u8 state = (*vector & DCCP_ACKVEC_STATE_MASK) >> 6;
	325	const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
	326
84116716	327	dccp_pr_debug_cat("%d,%d\|", state, rl);
ae31c339 ACM	328	++vector;
	329	}
	330
84116716	331	dccp_pr_debug_cat("\n");
ae31c339 ACM	332	}
	333
	334	void dccp_ackvec_print(const struct dccp_ackvec *av)
	335	{
	336	dccp_ackvector_print(av->dccpav_buf_ackno,
	337	av->dccpav_buf + av->dccpav_buf_head,
	338	av->dccpav_vec_len);
	339	}
	340	#endif
	341
02bcf28c AB	342	static void dccp_ackvec_throw_record(struct dccp_ackvec *av,
02bcf28c AB	343	struct dccp_ackvec_record *avr)
ae31c339	344	{
02bcf28c AB	345	struct dccp_ackvec_record *next;
02bcf28c AB	346
23d06e3b AB	347	/* sort out vector length */
	348	if (av->dccpav_buf_head <= avr->dccpavr_ack_ptr)
	349	av->dccpav_vec_len = avr->dccpavr_ack_ptr - av->dccpav_buf_head;
	350	else
	351	av->dccpav_vec_len = DCCP_MAX_ACKVEC_LEN - 1
	352	- av->dccpav_buf_head
	353	+ avr->dccpavr_ack_ptr;
02bcf28c AB	354
	355	/* free records */
	356	list_for_each_entry_safe_from(avr, next, &av->dccpav_records,
	357	dccpavr_node) {
	358	list_del_init(&avr->dccpavr_node);
	359	dccp_ackvec_record_delete(avr);
	360	}
ae31c339 ACM	361	}
	362
	363	void dccp_ackvec_check_rcv_ackno(struct dccp_ackvec av, struct sock sk,
	364	const u64 ackno)
	365	{
02bcf28c	366	struct dccp_ackvec_record *avr;
ae31c339	367
02bcf28c AB	368	/*
	369	* If we traverse backwards, it should be faster when we have large
	370	* windows. We will be receiving ACKs for stuff we sent a while back
	371	* -sorbo.
	372	*/
	373	list_for_each_entry_reverse(avr, &av->dccpav_records, dccpavr_node) {
	374	if (ackno == avr->dccpavr_ack_seqno) {
09dbc389	375	dccp_pr_debug("%s ACK packet 0, len=%d, ack_seqno=%llu, "
02bcf28c	376	"ack_ackno=%llu, ACKED!\n",
09dbc389	377	dccp_role(sk), 1,
02bcf28c AB	378	(unsigned long long)avr->dccpavr_ack_seqno,
	379	(unsigned long long)avr->dccpavr_ack_ackno);
	380	dccp_ackvec_throw_record(av, avr);
	381	break;
d23ca15a AB	382	} else if (avr->dccpavr_ack_seqno > ackno)
d23ca15a AB	383	break; /* old news */
ae31c339 ACM	384	}
	385	}
	386
	387	static void dccp_ackvec_check_rcv_ackvector(struct dccp_ackvec *av,
	388	struct sock *sk, u64 ackno,
	389	const unsigned char len,
	390	const unsigned char *vector)
	391	{
	392	unsigned char i;
02bcf28c	393	struct dccp_ackvec_record *avr;
ae31c339 ACM	394
ae31c339 ACM	395	/* Check if we actually sent an ACK vector */
02bcf28c	396	if (list_empty(&av->dccpav_records))
ae31c339	397	return;
ae31c339 ACM	398
ae31c339 ACM	399	i = len;
02bcf28c AB	400	/*
	401	* XXX
	402	* I think it might be more efficient to work backwards. See comment on
	403	* rcv_ackno. -sorbo.
	404	*/
	405	avr = list_entry(av->dccpav_records.next, struct dccp_ackvec_record,
	406	dccpavr_node);
ae31c339 ACM	407	while (i--) {
	408	const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
	409	u64 ackno_end_rl;
	410
	411	dccp_set_seqno(&ackno_end_rl, ackno - rl);
	412
	413	/*
02bcf28c AB	414	* If our AVR sequence number is greater than the ack, go
02bcf28c AB	415	* forward in the AVR list until it is not so.
ae31c339	416	*/
02bcf28c AB	417	list_for_each_entry_from(avr, &av->dccpav_records,
	418	dccpavr_node) {
	419	if (!after48(avr->dccpavr_ack_seqno, ackno))
	420	goto found;
	421	}
	422	/* End of the dccpav_records list, not found, exit */
	423	break;
	424	found:
	425	if (between48(avr->dccpavr_ack_seqno, ackno_end_rl, ackno)) {
8e27e465	426	const u8 state = *vector & DCCP_ACKVEC_STATE_MASK;
ae31c339	427	if (state != DCCP_ACKVEC_STATE_NOT_RECEIVED) {
09dbc389	428	dccp_pr_debug("%s ACK vector 0, len=%d, "
ae31c339 ACM	429	"ack_seqno=%llu, ack_ackno=%llu, "
ae31c339 ACM	430	"ACKED!\n",
09dbc389	431	dccp_role(sk), len,
ae31c339	432	(unsigned long long)
02bcf28c	433	avr->dccpavr_ack_seqno,
ae31c339	434	(unsigned long long)
02bcf28c AB	435	avr->dccpavr_ack_ackno);
02bcf28c AB	436	dccp_ackvec_throw_record(av, avr);
afec35e3	437	break;
ae31c339 ACM	438	}
ae31c339 ACM	439	/*
02bcf28c AB	440	* If it wasn't received, continue scanning... we might
02bcf28c AB	441	* find another one.
ae31c339	442	*/
ae31c339	443	}
ae31c339 ACM	444
	445	dccp_set_seqno(&ackno, ackno_end_rl - 1);
	446	++vector;
	447	}
	448	}
	449
	450	int dccp_ackvec_parse(struct sock sk, const struct sk_buff skb,
	451	const u8 opt, const u8 *value, const u8 len)
	452	{
	453	if (len > DCCP_MAX_ACKVEC_LEN)
	454	return -1;
	455
	456	/* dccp_ackvector_print(DCCP_SKB_CB(skb)->dccpd_ack_seq, value, len); */
	457	dccp_ackvec_check_rcv_ackvector(dccp_sk(sk)->dccps_hc_rx_ackvec, sk,
	458	DCCP_SKB_CB(skb)->dccpd_ack_seq,
	459	len, value);
	460	return 0;
	461	}
9b07ef5d	462
9b07ef5d ACM	463	int __init dccp_ackvec_init(void)
	464	{
	465	dccp_ackvec_slab = kmem_cache_create("dccp_ackvec",
	466	sizeof(struct dccp_ackvec), 0,
	467	SLAB_HWCACHE_ALIGN, NULL, NULL);
02bcf28c AB	468	if (dccp_ackvec_slab == NULL)
	469	goto out_err;
	470
	471	dccp_ackvec_record_slab =
	472	kmem_cache_create("dccp_ackvec_record",
	473	sizeof(struct dccp_ackvec_record),
	474	0, SLAB_HWCACHE_ALIGN, NULL, NULL);
	475	if (dccp_ackvec_record_slab == NULL)
	476	goto out_destroy_slab;
9b07ef5d ACM	477
9b07ef5d ACM	478	return 0;
02bcf28c AB	479
	480	out_destroy_slab:
	481	kmem_cache_destroy(dccp_ackvec_slab);
	482	dccp_ackvec_slab = NULL;
	483	out_err:
59348b19	484	DCCP_CRIT("Unable to create Ack Vector slab cache");
02bcf28c	485	return -ENOBUFS;
9b07ef5d ACM	486	}
	487
	488	void dccp_ackvec_exit(void)
	489	{
	490	if (dccp_ackvec_slab != NULL) {
	491	kmem_cache_destroy(dccp_ackvec_slab);
	492	dccp_ackvec_slab = NULL;
	493	}
02bcf28c AB	494	if (dccp_ackvec_record_slab != NULL) {
	495	kmem_cache_destroy(dccp_ackvec_record_slab);
	496	dccp_ackvec_record_slab = NULL;
	497	}
9b07ef5d	498	}